Zurich Open Repository and

Archive
University of Zurich
University Library
Strickhofstrasse 39
CH-8057 Zurich
www.zora.uzh.ch

Year: 2022

The role of language documentation in corpus-based typology

Schnell, Stefan ; Haig, Geoffrey ; Seifart, Frank

Posted at the Zurich Open Repository and Archive, University of Zurich

ZORA URL: https://doi.org/10.5167/uzh-217097
Book Section
Published Version

The following work is licensed under a Creative Commons: Attribution-NonCommercial 4.0 International (CC
BY-NC 4.0) License.

Originally published at:

Schnell, Stefan; Haig, Geoffrey; Seifart, Frank (2022). The role of language documentation in corpus-based typol-
ogy. In: Haig, Geoffrey; Schnell, Stefan; Seifart, Frank. Doing Corpus-Based Typology With Spoken Language
Corpora: State of the art. Honolulu, HI, U.S.A.: University of Hawai’i Press, 1-28.
 Language Documentation & Conservation Special Publication 25 1
Doing corpus-based typology with spoken language data: State of the art
edited by Geoffrey Haig, Stefan Schnell, and Frank Seifart, pp. 1 – 28
https://nflrc.hawaii.edu/ldc/sp25
https://hdl.handle.net/10125/74656

The role of language documentation

in corpus-based typology

Stefan Schnell,1 Geoffrey Haig,2 Frank Seifart 3

1 2 3
University of Zurich, University of Bamberg, Leibniz-Centre General Linguistics (ZAS)

1 Introduction
The global decline in linguistic diversity was first brought to public atten-
tion some 30 years ago (Hale et al. 1992), and has since continued unabated
(Seifart et al. 2018; Bromham et al. 2021). From the perspective of language
communities, language loss means an irredeemable rupture of collective and
individual memory; for the language sciences, each disappearing language
shrinks our window on the range of variability in human language. The ur-
gency of language documentation could scarcely be more obvious.1
For typology, a major utility of language documentation was the adequate
representation of typological rara in linguistic theory: syntactic ergativity,

1 This volume grew out of a workshop on Corpus-based Typology: Spoken Language from a
Cross-linguistic Perspective, held at the Annual General Meeting of the German Linguistics
Society (DGfS) at the University of Hamburg in March 2020. We, the editors, would like
to express our thanks to the audience at the workshop for very stimulating discussion, to
the external reviewers of the contributions of this volume for their critical feedback, and
to Nils Schiborr for supervising the final volume production. The responsibility for any
remaining errors is of course our own.
2 LD&C SP25 — Doing corpus-based typology with spoken language data

OSV constituent order, non-configurationality, and so on (Comrie 1981: 5–

12). But more recently, the role of language documentation in typology has
been re-defined by a shift towards a more usage-based approach to typology,
incorporating methodological advances from corpus linguistics, variationist
sociolinguistics, and cognitive sciences. Rather than extracting linguistic gen-
eralizations from pre-processed data as provided by descriptive grammars,
the primary data created by language documentation itself, in the form of re-
cords of language usage, can provide the input for a usage-based approach to
typology which seeks to go beyond the emphasis of traditional typology on
construction types outside of context (e.g. Greenberg 1963), and incorporate
patterns of variation and the impact of context.
Although it was not necessarily foreseen by earlier documentary linguist-
ics, it transpires that the emphasis on maximally context-embedded docu-
mentation of language usage yields precisely the kind of data that is amenable
for usage-based approaches to typology. This is particularly noteworthy in
view of the critique of this kind of documentation as a “data graveyard” (New-
man 2013). Data graveyards are now data goldmines, and typologists are just
beginning to explore their potential (Seifart et al. 2012). In this volume we
showcase a selection of state-of-the art research that embodies this usage-
and corpus-based approach to typology, drawing extensively (but not exclus-
ively) on primary linguistic data compiled and archived initially for the pur-
pose of language documentation. In the remainder of this introduction, we
outline in Section 2 a research agenda which we refer to as ‘corpus-based ty-
pology’ (CBT), outlining influential developments in corpus linguistics and
typology, recent advances and remaining challenges. In Section 3 we sum-
marize in what ways the contributions aggregated in this volume address
some of these challenges, before concluding in Section 4.

2 Corpus-based typology
Corpus-based typology refers to a set of approaches that use corpora to con-
duct language typology. The following sections are devoted to key aspects of
this research agenda: Section 2.1 addresses the concept of ‘corpus’ in a cross-
linguistic context; Section 2.2 looks at the role of language-internal variation;
Section 2.3 addresses different aspects of language usage. We refer the reader
Language documentation and corpus-based typology 3

to Levshina (2021a) and Schnell & Schiborr (in press) for more extensive over-
views of recent developments; here we merely summarize the main points
before identifying remaining challenges and desiderata (Section 2.4), which
the current volume intends to address.

2.1 Multilingual corpora

CBT deploys methods developed in neighbouring disciplines, such as corpus
linguistics and variationist sociolinguistics (see below), to address a range
of questions relevant for language typology. Hence, the research showcased
in this volume all draw on corpora, understood here as follows (cf. Barth &
Schnell 2021: 2–8): A corpus is a text or a collection of texts which are in turn
defined as a succession of connected utterances that form a coherent whole.
Texts in a corpus can be spoken, signed, or written. An important aspect of
corpora is that individual linguistic structures and texts as a whole can be
analysed with regards to its contextual determinants, although in practice
the extent to which such analyses are possible depends heavily on the level
of linguistic representation and their immediate context (e.g. sound segment
within a phonological word or sentence within a text) and the detail and
quality of metadata (for entire texts). These features differentiate a corpus
from collections of elicited, context-free sentences of the kind frequently used
in grammars, and from paradigms and word lists. A corpus in this sense
permits the analysis of, for example, clause-combining, prosodic chunking,
referent tracking, turn-taking, narrative structure, and so on. To serve as
the input to corpus-based typology, a corpus in this sense should provide
sufficient numbers of tokens relevant to a given research question to enable
the identification of relevant patterns, possibly through meaningful statistical
analysis.
Corpora in typology differ along a scale of content control, that is the
degree to which the content of the corpus is pre-determined by the investig-
ator. At the highest end of the scale are parallel texts: Corpora produced as
translations from a single source text, for example the Parallel Bible Corpus
(PBC) (Mayer & Cysouw 2014) and the Universal Declaration of Human Rights
4 LD&C SP25 — Doing corpus-based typology with spoken language data

(UDHR)2 for which translations in about 1400 languages are now available
(cf. Bentz & Ferrer-i-Cancho 2016). The VoxClamantis project (Salesky et al.
2020) is a corpus of Bible translations read aloud by native speakers, with first-
pass phoneme-level alignments for more than 600 languages. A lower degree
of content-control is found in so-called parallax corpora, where speakers pro-
duce texts in response to a specific non-verbal stimulus. The best-known ex-
amples are those based on re-tellings of the Pear Story video (Chafe 1980),
or the Frog Story picture book (Mayer 1969; see Berman & Slobin 1994). See
Barth & Evans (2017) and Barth & Evans (this volume) for examples of the
parallax methodology. Finally, at the lowest level of content control are cor-
pora produced without any specific pre-defined content constraints. These
include, for example, life stories, traditional narratives, descriptions of activ-
ities, and so on, loosely referred to as “original text typology” (Haig et al.
2011). Texts of this type are among the most common outputs of language
documentation, and are thus of particular relevance in the current connec-
tion. Degree of content control has a considerable impact on the nature of
the typological questions that can be addressed: High content control is de-
signed to elicit a high degree of semantic consistency across different cor-
pora, and is thus ideal for probing, for instance, cross-linguistic differences
and commonalities of specific event types in specific contexts (e.g. motion
events, Wälchli & Sölling 2013). However, high content control comes at the
cost of a lack of naturalness, and possible interference from the source texts.
Corpora also vary across other dimensions, for example medium (spoken,
signed, or written, further discussed in Section 2.4 below), single-participant
(monologue) or multi-participant, register, and genre. This volume is de-
voted to spoken language corpora, which always minimally involve record-
ing, and generally transcription, followed by further levels of annotation,
which may exhibit various degrees of conceptual abstraction, from a simple,
content-based translation to exhaustive morpheme-for-morpheme glossing,
with indications of prosody, information structure, or semantic roles. See Sei-
fart (this volume), Mettouchi & Vanhove (this volume) and Haig et al. (this
volume) for different approaches to phonological, prosodic, morphosyntactic,
and information structure annotation.

2 https://www.unicode.org/udhr/
Language documentation and corpus-based typology 5

Common to all these annotation schemes is the commitment to maximal

cross-linguistic applicability of the annotated categories. In this respect, CBT
on spoken language corpora aligns with recent approaches to corpus- or
token-based typology based on Universal Dependencies (UDs), a widely-used
system of morpho-syntactic annotation that is primarily implemented semi-
automatically on standardized written corpora (Zeman et al. 2021)3 . Both
compare corpora of different languages, both aim at deriving higher-level
generalizations extracted directly from the corpora, rather than via the medi-
ation of pre-formulated grammatical analysis (see Wälchli 2009 on the latter
point), and both develop annotation schemes specifically designed to be ap-
plicable across a typologically diverse range of languages.

2.2 Variation centre stage

Typology in the tradition of Greenberg’s seminal work focused on identify-
ing limits of cross-language variation by classifying languages according to
construction types (e.g. “prepositional”, “SOV”, etc.). Sampling procedures
were developed to ensure that observed (dis)preferences were really univer-
sal to human language rather than stemming from shared histories of inher-
itance (i.e. from a genealogical bias), or from extensive long-term contact (i.e.
from an areal bias). Explanations for (statistical) limits of variation were then
sought in universal properties of human communication and language pro-
cessing. For instance, the near-absence of OSV in the world’s languages is
explained in terms of violating two functional principles, “S before O” (new
before given), and also “O should be adjacent to V”, due to the close concep-
tual link between objects and verbs (e.g. Tomlin 1986; Hawkins 1994).
However, this “whole-language” typological approach glosses over a lot
of the intra-linguistic variation (Bickel 2009). For instance, languages may
show different word order rules in different sentence types (cf. independent
and VO vs. subordinate clauses and OV in German) or different alignment
systems (e.g. split in terms of tense, as in Iranian languages, or split along
different encoding devices, e.g. ergative case marking and accusative agree-
ment in Tukang Besi; Donohue 1999: 51–54). Zooming in even closer to intra-

3 https://universaldependencies.org/
6 LD&C SP25 — Doing corpus-based typology with spoken language data

linguistic variation, Multivariate Typology (MVT, Bickel 2015) suggests that

linguists should focus on low-level aspects of language, using what Levin-
son & Evans (2010: 2738–2739) call Low-Level Feature Metalanguage (LLFM).
This low-level approach advocates late aggregation, where the investigation
of low-level categories precedes the identification of higher-level concepts
(Zakharko et al. 2017). In this bottom-up approach, typological and poten-
tially universal trends are statistically significant clusters of properties. Cru-
cially, variation is not glossed over in MVT, since observations that do not
follow a given trend are still represented as statistical minorities or outliers.
The concern of CBT with variation builds on well-established founda-
tions, most notably in variationist sociolinguistics in the tradition of Wil-
liam Labov (1972, 1994), anthropological linguistics (e.g. Hymes 1961, 1962;
Duranti 1981, 1997), conversation analysis (Schegloff 2006; Sacks et al. 1974),
research in language acquisition (MacWhinney 2000; Slobin 1985 and later
works) as well as general corpus linguistics (McEnery & Wilson 2001 among
many others). More recently, these sub-disciplines have moved beyond study-
ing a small number of well-researched languages to investigating variation
across increasingly diverse language samples; see the contributions in Stan-
ford & Preston (2009), and Mansfield & Stanford (2017) on sociolinguistic
studies of understudied languages; Meyerhoff (2009) and Torres Cacoullos &
Travis (2018) on corpus-based approaches to null-subjects across languages;
and Dingemanse et al. (2014) as well as Dingemanse et al. (2017) for prag-
matics, including conversation analysis. Similarly, in more general corpus
linguistics, studies in register and genre variation have also been conducted
on lesser-studied languages (cf. Schnell & Barth 2018) and in comparative
perspective (Biber 1995).

2.3 Language use

Despite its concern for intra-linguistic variation, MVT is primarily concerned
with language systems and different structural subtypes. CBT research, on
the other hand, aims at capturing intra-linguistic variation in language use
and its relation to aspects of language systems. It may thus, for example,
recast word order typology by refining characterizations such as “dominant
OV” (Dryer 1992) as “73% OV” (Gerdes et al. 2021). Moreover, it can advance
conceptions of linguistic categories as clusters of contextualized exemplars
Language documentation and corpus-based typology 7

in language use. For instance, Cysouw (2014) induces semantic roles (such as
Goal, Agent, or Patient), or even macro-roles (like S, A, and P) from the usage
of case-like markers across 15 languages, based on parallel religious texts. In
this section we give examples of research that seeks motivations for typolo-
gical (dis)preferences in different levels of linguistics structure in patterns
of language use. These examples illustrate what CBT’s corpus-linguistic
approach adds to the study of variability beyond the recognition of intra-
language variation that is already registered within MVT and related ap-
proaches, which can, in principle, be applied to, for example, elicited or pre-
processed data.
In syntactic typology, dependency length minimisation (DLM) has long
been hypothesized to underlie the typological preference for harmonic word
order patterns, for instance VO co-occurring with prepositions in head-initial
languages (cf. Behaghel 1909; Dryer 1992; Hawkins 2014, 2004). How DLM
shapes actual language use has now been directly investigated comparatively
in corpora from over 50 languages (Futrell et al. 2020; Jing et al. 2021). Simil-
arly, there are now corpus-based results on the avoidance of crossing depend-
encies (Blasi et al. 2019), as well as cross-linguistic differences in word order
variability (more or less “free” word order) (Futrell et al. 2015; Levshina 2019)
and the trade-off between word order variability and case marking (Koplenig
et al. 2017; Levshina 2019, 2021b).
Another classic topic in syntactic typology is zero reference, discussed
in the theoretical and typological literature under the heading of “pro-drop”
(Rizzi 1982; Jaeggli & Safir 1989; Roberts & Holmberg 2009; cf. Dryer 2013).
Bickel (2003) and Stoll & Bickel (2009) investigate the rate of zero expression
of syntactic arguments across a set of three languages from the Himalayas,
and Russian. Factors impacting on referential density, that is the rate of overt
to covert forms of reference, are identified as patterns of clause combining
as well as ethnolinguistic considerations of discourse production and recep-
tion. While these authors focus on possible realizations of all arguments in
a particular corpus, Torres Cacoullos & Travis (2019) focus on the specific
conditions of zero reference for subjects, finding, for instance, that the same
factors are involved in both English and Spanish texts, but differ in their de-
gree of magnitude. Vollmer (2019) applies classification tree methodologies
to investigate referential choice in a diverse sample of 8 languages, likewise
identifying underlying commonalities across diverse corpora. These works
8 LD&C SP25 — Doing corpus-based typology with spoken language data

thus open up comparative and quantitative ways to address systematic dif-

ferences in discourse production that can be correlated with other linguistic
features and/or cultural differences.
In morphological typology, marking asymmetries such as the presence
versus absence of morphological marking on word pairs like Spanish romper
‘(causative) break’ and romper-se ‘(inchoative) break’, have long been dis-
cussed (e.g. Greenberg 1966). Haspelmath et al. (2014; see also Haspelmath
2021) found that, across corpora from various languages, overt morpholo-
gical marking corresponds to lower mention frequencies, compared to the
unmarked member. This provides a motivation for marking asymmetries
in human language based on information-theoretic accounts in terms of ef-
ficiency going back to Zipf (1935; cf. Gibson et al. 2019 for an overview).
Regarding word length more generally, recent work in CBT has found sup-
port for Zipf’s Law of Abbreviation, which describes the relation between
the length of words and their text frequencies in corpora from about 1400
languages (Bentz & Ferrer-i-Cancho 2016). Other recent CBT research found
that word length is even more closely related to a higher predictability of
words in context, compared to the raw text frequency of a word (Piantadosi
et al. 2011; Gibson et al. 2019). Similarly, albeit more directed towards system
(lexicon-related) properties, is Cohen Priva’s (2017) study of form reduction
and functional load that essentially finds that reduction is more likely with
elements that have a low functional load.
In pioneering work at the interface of prosody and morphological typo-
logy, Himmelmann (2014) studied patterns of hesitations and disfluencies ob-
served in spoken corpora of English, German, and Tagalog, providing an al-
ternative explanation for the suffixing preference in the world’s languages
(Cutler et al. 1985; Bybee et al. 1990). Word class-specific hesitation pat-
terns were studied in Seifart et al. (2018) across nine typologically diverse
languages. Both studies intend to link typological preferences in the mor-
phology of the world’s languages to speech and sentence planning by taking
speech rate reduction and pauses, as observable in spoken corpora, as a proxy
indicating planning difficulties.
Focussing on potentially universal characteristics of information trans-
mission in spoken language production, Pimentel et al. (2021) study the dur-
ation of phones as a function of surprisal across 600 languages, based on the
VoxClamantis corpus, finding evidence for a surprisal-duration trade-off, and
Language documentation and corpus-based typology 9

thus for the Uniform Information Density hypothesis (Frank & Jaeger 2008).
Comparing 17 languages, Coupé et al. (2019) find that higher syllable com-
plexity (which induces higher informativity of each syllable) correlates with
slower articulation rate, resulting in a pattern whereby information trans-
mission rates across languages are surprisingly comparable. Regarding the
temporal chunking of speech, Inbar et al. (2020) measured the regularity of
sequences of intonation units across six spoken corpora, with results suggest-
ing that they closely match brain waves at 1Hz.

2.4 Remaining challenges

Significant results are already emerging from CBT studies that connect pref-
erences in usage and underlying processing biases much more immediately
to typological distributions of linguistic structures. Remaining challenges re-
late to, firstly, reliance on written modality, secondly, relatively small and
biased samples in terms of languages and genres, and, thirdly, the comparab-
ility of the corpora regarding contents and annotation.
The large collections of Universal Dependency-annotated texts, and the
large parallel text corpora such as the PBC consist predominantly or exclus-
ively of published written texts. These typically involve multiple passes of
editing (potentially by multiple parties). UD corpora and the PBC are thus
fairly remote from the primary, spoken or signed, modes of human speech
production and reception. A recent example highlighting these issues is Just
& Čéplö’s (to appear) corpus-based investigation of bound object indexing
in Maltese, an apparently widespread phenomenon in spoken Maltese. Their
initial attempt to work with UD corpora of Maltese had to be abandoned
because of the scarcity of the relevant constructions in the written corpora.
Though non-comparative, Just & Čéplö’s study suggests that considerations
of mode will likely be highly relevant for comparative work in CBT as well.
Further initial support for this comes from Schnell & Schiborr’s (in press)
comparison of (predominantly written) texts in UDs and the entirely spoken
texts from Multi-CAST (Haig & Schnell 2021) which suggests that consider-
ations of DLM may be hardly relevant in the latter corpus simply because
it lacks critical proportions of complex, long NPs. Relatedly, greater toler-
ance for more, and more complex NPs in written texts has been explained
by the greater ease of written text comprehension. Thus, written text pro-
10 LD&C SP25 — Doing corpus-based typology with spoken language data

duction represented in UDs does not directly reflect relevant constraints on

processing themselves, although effects of such constraints may be carried
over from spoken or signed text production. Generally, spoken or signed text
production is of primary interest to CBT since only here do we find consid-
erations of language processing and efficiency-related trade-offs playing out
under relevant time constraints of production and comprehension as well as
noisy channel and similar considerations of the articulation bottleneck, prin-
ciples of inference, and so forth.
In addition to modes of production and reception, current efforts in CBT
are fairly constrained in the groups of language users they represent. On
a global scale, this concerns the number and diversity of languages: While
the PBC and VoxClamantis collections include subcorpora from a number of
understudied languages (Bible translations are often prepared as part of mis-
sionary linguistic work in small language communities), UD corpora show a
massive imbalance towards better-studied, often standard written languages
of western Europe, largely corresponding to speaker populations identified as
WEIRD and non-representative of humanity more generally by Henrich et al.
(2010). And both the European and the few non-European languages repres-
ented in UD corpora are mostly those that Dahl (2015) labels LOL languages
(“Literary tradition, Official language status, Large number of speakers”), like-
wise a tiny minority among the world’s languages, and demonstrably not
representative of these.
Likewise, intralinguistic variation in terms of, for example, speakers,
genres, and registers is not properly reflected in current CBT research and
cross-linguistic corpora: An extreme example are the PBC and VoxClamantis,
which both represent a single text for each language community that is relev-
ant for an extremely limited social register (practice of a religious cult), often
excluding large proportions of any given community who do not read their
language fluently. Though variation within understudied languages has only
recently gained more attention (cf. Hildebrandt et al. 2017; Mansfield & Stan-
ford 2017 and references therein), it seems obvious from the long tradition
of sociolinguistics that inter-user variation is universal across languages and
needs to be included in CBT (cf. Barth et al., this volume). Concerning vari-
ation in terms of text types, Schnell & Barth’s (2018) study of object pronoun
use in the Oceanic language Vera’a points to the relevance of considering
cross-register variation for their finding of discourse topicality (rather than
Language documentation and corpus-based typology 11

animacy) as the most relevant factor of pronoun use. Biber (1995: 359) finds
differences across registers to be remarkably similar across diverse languages
(English, Somali, Korean, Nukulaelae Tuvaluan),4 pointing to the high rel-
evance of considering cross-register variation in cross-linguistic research of
language use.
Finally, CBT has to confront the problem of the relative comparability of
corpora: Bickel (2003) makes a strong case for the use of content-controlled
data in a study of referential density, given the high content-sensitivity of
the parameter under study. Given the lesser representativeness of such ex-
perimentally elicited texts, an alternative approach is to determine pragmat-
ically defined usage contexts that can be considered broadly comparable, as
advanced in pragmatic typology (e.g. Dingemanse et al. 2014). Which of these
two approaches is best suited for a given research agenda will depend heavily
on the respective projects.

3 CBT and language documentation: Case

studies
In the following sections we identify four areas where CBT and language
documentation are already yielding a fertile union, addressing some of the
challenges in areas of cross-linguistic and typological research identified in
the previous section. We illustrate each of these areas with reference to the
contributions to this volume. Section 3.1 looks at language acquisition in
language documentation, Section 3.2 at the interface between prosody and
morpho-syntax, Section 3.3 at cross-linguistic approaches to interaction, in-
formation packaging, animacy, and morpho-syntax, while Section 3.4 reports
on cross-linguistic variation vis-à-vis intra-speaker variability in social cog-
nition. Lastly, Section 3.5 discusses the challenges of corpus processing, an-
notation, and usability.

4 Biber (1995) focuses on a comparative study of English and Somali registers, yet the assess-
ment encompasses all four languages.
12 LD&C SP25 — Doing corpus-based typology with spoken language data

3.1 Acquisition in language documentation (Hellwig et al.)

In research on language acquisition corpus-based work has a long tradi-
tion (MacWhinney 2000). But availability of corpora of children’s language
production and child-directed speech are to date still limited to much less
than 2% of the world’s languages (Stoll & Bickel 2013: 198), and biased to-
wards WEIRD societies and LOL languages. Efforts to broaden the scope
of acquisition corpora have recently gained traction, for instance within the
ACQ(uisition )DIV(ersity) research group led by Sabine Stoll at the Univer-
sity of Zurich (Moran et al. 2016). Hellwig et al. (this volume) outline the
Sketch Acquisition project, which addresses the challenges of including a
range of underdocumented and understudied languages in corpus building
for language acquisition research while still compiling sufficient data for ro-
bust findings on acquisition. They stress the importance of an open corpus
design for this kind of research matching the spirit of language documenta-
tion by keeping corpora maximally amenable to a range of research questions
and projects. This also means that corpus annotations should be kept at a re-
latively general level, capturing what is relevant for a specific agenda, but
leaving open the pursuit for other types of research. The authors report on a
range of first-hand novel observations of children’s verbal behaviour during
language acquisition in the Papuan language Qaqet. For example, in their Sec-
tion 2.2.1 the authors note that in this speech community, children’s verbal
interactions are mostly with other children, and these child-to-child interac-
tions are characterized by a high frequency of exact repetitions. These find-
ings serve as an important corrective to conventional views on the nature of
the acquisition process, largely based on data from a small number of WEIRD
speech communities, with a focus on adult-to-child transmission.

3.2 The interface of prosody and morpho-syntax: Towards a

typologically informed and corpus-based approach (Seifart;
Mettouchi & Vanhove)
As mentioned in Section 2.4 above, the study of specific properties of spoken
language discourse production is a central concern for CBT, as this is the ac-
tual arena where discourse processing, prosodic chunking, and the formation
of linguistic units plays out. Seifart (this volume) reports on research into
Language documentation and corpus-based typology 13

durational prosodic properties of spoken text from diverse languages based

on the multilingual corpus DoReCo (“Language DOcumentation REference
COrpus”). DoReCo’s goal is to significantly enhance the coverage of global
linguistic diversity in spoken language corpus data, aiming at inclusion of 50
language corpora in the first release planned for mid-2022. For that purpose,
DoReCo selects and processes subsets of language documentation collections,
consisting mostly of traditional and personal narratives. Its particular value
for linguistic analysis is the phonetic segment-level time-alignment of the
transcription and dependent annotations, including morphological glossing
for 35 languages.
Vanhove and Mettouchi’s (this volume) contribution showcases two col-
lections of multilingual corpora, collectively compiled and annotated: CorpA-
froAs (2007–2012) and CorTypo (2013–2017). The first focuses on thirteen
Afro-Asiatic languages, while the latter includes twelve languages from a
range of phyla and interfaces the corpora with a searchable typological data-
base. Both are characterized by an exceptionally rich annotation scheme, im-
plemented in a purpose-built variant of the ELAN software (ELAN developers
2020; Chanard 2015) and combining morphological, syntactic, and prosodic
segmentation. Their morphosyntactic annotation draws on those categor-
ies and features which are demonstrably relevant in the actual data, through
“slow empirical building of established language-internal categories as a basis
for further comparison, in a bottom-up perspective” (Mettouchi & Vanhove,
this volume; see Section 2 above). Another crucial feature is the systematic
segmentation of the corpora into intonation units (Izre’el & Mettouchi 2015).
The result is an extremely fine-grained glossing and the possibility of detailed
cross-corpus queries using the CorpAfroAs and CorTypo websites. Several
case studies are presented illustrating the potential of this approach. One
of these finds hitherto unnoticed interactions between word order and the
prosodic integration of reported speech across a sample of four languages
(Beja, Zaar [Chadic], Juba Arabic [Creole], and Modern Hebrew [Semitic]):
In SOV languages, the onset of the speech report is systematically set off
from the previous intonation unit through a clear prosodic cue whereas in
SVO languages, it is the end of the speech report which is set off from the
next intonation unit.
14 LD&C SP25 — Doing corpus-based typology with spoken language data

3.3 Cross-linguistic approaches to interaction, information

structure, animacy, and morphosyntax (Ozerov; Haig et al.)
Another area of discourse production concerns the structuring and delivery
of information: Producers typically have to navigate between their own plan-
ning and production efforts while considering the needs of their addressees
and how to draw their attention to what is of most interest at any given point
in discourse. Studies that address such questions are often confined to single
languages and have been relying to a major extent on macro-categories like
topic and focus, which upon closer inspection have been found to be not
readily associable with specific means of expression across languages (Matić
& Wedgwood 2013; Ozerov This volume, among others).
An alternative approach is taken by Ozerov (This volume), who conceives
of information packaging as a level of structuring that is derivative of the
clustering of distinct interactional moves in discourse production. Within
this framework, Ozerov (this volume) investigates question formation and
so-called detachment constructions together with a range of concomitant
structural aspects, including prosodic aspects. Ozerov demonstrates that in-
dividual structures are more readily associated with specific interactional
moves (e.g. getting addressee’s attention through pitch) and/or addressee-
oriented effects (e.g. pausing after mentioning an entity) that may be inter-
preted by addressees in certain ways (e.g. this entity is relevant for the im-
mediately subsequent discourse). Comparing two unrelated languages, Anal
Naga (Tibeto-Burman) and Ivrit (also known as Modern Israeli Hebrew), this
study suggests that the low-level categories of individual interactional moves
find similar expressions across different languages and are also a suitable
basis for cross-linguistic comparisons.
A further aspect of information packaging relates to argument marking
and argument expression, that is the alternation between full NPs versus pro-
nouns and zero reference. Haig et al. (this volume) investigate the rates of full
NPs across corpora from 15 languages, finding astonishingly stable rates of
full lexical NPs across all corpora. This is noteworthy given that the corpus
texts have not been controlled for content, unlike Bickel’s (2003) or Stoll and
Bickel’s (2009) parallax corpora. Furthermore, this finding suggests an over-
all consistent response by discourse producers to motivations of informative-
ness and tellability on the one hand (pushing up the rate of lexical NPs) and to
Language documentation and corpus-based typology 15

discourse coherence and referential continuity on the other hand (dragging

down the rate of NPs).
The authors also investigate the related question of how full NPs are
marked for their syntactic role, specifically the differential marking of NPs
for different syntactic functions as an example of marking asymmetries, fo-
cusing here on the functions A and P in transitive clauses. The authors assess
earlier accounts of differential argument marking in terms of frequencies of
verbalization of participant role with different semantic properties (person,
animacy, definiteness; cf. Haspelmath 2021). Their study suggests that while
attested corpus frequencies do support the overall trend to overtly mark the
less frequent member of a particular opposition, the results do not square
up with the attested frequencies for different systems of differential marking
of A and P in the world’s languages, in particular the comparative rarity of
differential A marking. For this, considerations of ambiguity in connection
with efficiency, as explicated by Piantadosi et al. (2012), may be more relevant
than corpus frequency.

3.4 Investigating social cognition: Purpose-built corpora (Barth

et al.)
An aspect of spoken and signed discourse production that goes beyond spe-
cific structures of the texts themselves is how language users navigate the
social reality they communicate in and about. This dimension of social cog-
nition is relevant for any discourse in any language; yet of particular interest
here is a comparison of languages and their use in diverse cultures, to find
out how different cultural backgrounds (concepts, values) influence the way
language users verbalize states of affairs in discourse. Also, matters of social
cognition are mostly relevant in face-to-face interactions.
These issues are addressed by the Social Cognition project that Barth et al.
(this volume) report on. Corpus-based research into social cognition across
diverse, mostly only spoken and signed languages runs into the problem of
finding sufficient data points that are comparable across languages, and are
relevant for the research questions at hand. For instance, free narrative cor-
pora normally do not contain many lexical NPs referring to humans since
these are often salient discourse referents that tend to be expressed by non-
lexical forms. Additionally, available cross-linguistic corpora rarely properly
16 LD&C SP25 — Doing corpus-based typology with spoken language data

cover potential variation across language users (as mentioned in Section 2.4).
Hence, to investigate lexical choices in reference to humans across diverse
languages, Social Cognition Parallax Interview Corpus (SCOPIC) texts are
elicited with the help of a picture task stimulus that is designed so as to elicit
instances of human references, and likewise lexical choices in other domains.
This optimally enables comparisons across languages and across individual
users, since contexts are kept stable. Barth et al. (this volume) present four
case studies on intra-language and intra-language user variation in a sample
of thirteen languages using these data. Overall, the authors find substan-
tial variation across users of a single language, which actually exceeds the
variability between languages, especially in the case of research questions
in semantic typology. These findings highlight the necessity for closer mon-
itoring of community-level representativeness in CBT and for investigating
methods that allow researchers to assess contribution of individual language
users.

3.5 Corpus processing, annotation, and usability

Our focus on spoken-language material from understudied languages comes
with the attendant burden of exceptionally labour-intensive processing and
annotation of data. Such data are typically recorded and annotated together
with native speakers during fieldwork, often in the absence of standardized
orthographies, reference sources such as dictionaries and grammars, or other
luxuries taken for granted in most of corpus linguistics. This is quite ob-
viously the major reason why corpora of spoken- or signed-language texts
from underdescribed languages remain relatively small. Moreover, only a
small number of research questions, such as studies of word length, can be
undertaken on the basis of transcriptions alone, while investigations of most
prosodic, morphosyntactic, and other structures usually necessitate further
annotation. All contributions in this volume detail the kind of corpus pro-
cessing and annotation involved in the corpus development projects they
build on, and we will not go into more detail here (cf. also Barth & Schnell
2021: Chapters 6 and 7).
A major issue in the further development of CBT is the balance between
diversity and standardization of annotation schemes across a scientific com-
munity with very diverse theoretical backgrounds working on very diverse
Language documentation and corpus-based typology 17

languages. Even Universal Dependency annotations, which have become a

fairly widely-accepted standard in syntactic parsing of written corpora, have
spawned a range of “dialectal” variants and offshoots. For example, Gerdes et
al. (2021) use the variant annotation system SUD (Surface-Syntactic Univer-
sal Dependencies), which implements the opposite direction of dependency
between adpositions and complements when compared to standard UD. The
GRAID system (Haig & Schnell 2014) posits hierarchical levels of annotation,
utilizing top-level categories such as ⟨pro⟩ (‘free definite pronoun’) as a re-
quired minimum, which can be (optionally) further specified, for instance
as ⟨rel_pro⟩ ‘relative pronoun’. The Hambam Corpus (Hamedan-Bamberg
Corpus of Contemporary Spoken Persian) implements a simplified version
of GRAID, GRAID-L,5 which is largely interoperable with standard GRAID,
and these examples could be multiplied.
Like corpus development more generally (cf. Hellwig et al., this volume),
the development of corpus annotation schemes can profit from adapting
the spirit of language documentation of keeping conventions as open, trans-
parently documented, and flexible as possible to be amenable to further re-
search agendas. While standards for phonemic transcriptions and the Leipzig
Glossing Rules for morphemic annotation seem to be firmly established, it is
still unclear whether in future developments of CBT further gold standards
of annotation will evolve, especially for higher structural levels.
Similar issues arise in relation to the comparability of texts and corpora
across languages and individual users. As discussed above, for some research
agendas more immediate comparability is key, for instance when it comes to
comparison of rates of zero reference or reference to humans. As with stand-
ards of corpus annotation, there is no gold standard regarding comparability
of this order. In fact, most studies in this volume draw on texts that have ori-
ginally been selected for recording for documentary reasons, including upon
request from community members, showing that such uncontrolled data is
amenable to a range of research questions (in the same way as likewise un-
controlled UD corpora are). While language documentation as conceived by
Himmelmann (1998) does encompass the collection of more controlled data
for specific descriptive and analytical purposes (though this has often fallen

5 https://multicast.aspra.uni-bamberg.de/resources/hambam/
18 LD&C SP25 — Doing corpus-based typology with spoken language data

by the wayside in the reception and citation of this seminal article), there are
good reasons from a collaborative documentary point of view to give preced-
ence to communities’ preferences, often for traditional narratives, and the
contributions in this volume bear witness to the fact that such data are most
useful for CBT. This is corroborated by a general corpus-linguistic view that
seeks to determine various conditions on language use, so that confinement
to experimentally elicited data would yield too narrow a scope of situational
factors.
Finally, corpora need to be sustainably and accessibly archived with clear
instructions how they can be used, in line with best practices of documentary
linguistics (Gippert et al. 2006; Thieberger & Berez 1963) and of open and
reproducible science (Wilkinson et al. 2016; Berez-Kroeker et al. 2018). Nearly
all contributors to this special publication are in the process of building web-
accessible corpora and hence contribute to the enterprise of open science
and collaborative research that CBT relies on much more than other fields in
linguistics.

4 Conclusions
Its initial success story notwithstanding, CBT is still in its infancy and unified
research agendas and standards are still emerging. Considerable advances
have already been made by researchers working primarily on digital corpora
of written standardized languages. This volume broadens the scope of CBT
by connecting it with language documentation, enabling a shift towards com-
parison based on context-embedded, naturally variable spoken and signed
language usage, where the social and cognitive-articulatory factors that mo-
tivate typological generalizations are actually operative. As coverage of the
world’s languages in spoken and signed corpora grows, CBT will also be able
to feed into a more holistic approach to areal and diachronic typology by in-
vestigating the conditioned use of linguistic structures by different users with
different demographics and across social settings and communities with dif-
ferent cultural backgrounds. This development will incur major challenges
in data collection and processing, including in particular linguistic annota-
tions of various kinds for comparative purposes. While this amounts to an
enormous undertaking, evoking a “sea change” in linguistics propagated by
Language documentation and corpus-based typology 19

Levinson & Evans (2010), we believe that the confluence of current advances
in corpus-based typology and language documentation will bring us a signi-
ficant step further in this direction.

References
Barth, Danielle & Evans, Nicholas. 2017. SCOPIC design and overview. In Barth,
Danielle & Evans, Nicholas (eds.), The Social Cognition Parallax Interview Cor-
pus (SCOPIC): A cross-linguistic resource (Language Documentation & Conservation
special publication 12), 1–23. Honolulu, HI: University of Hawai’i Press. (https:
//hdl.handle.net/10125/24742).
Barth, Danielle & Evans, Nicholas & Arka, I Wayan & Bergqvist, Henrik & Forker,
Diana & Gipper, Sonja & Hodge, Gabrielle & Kashima, Eri & Kasuga, Yuki
& Kawakami, Carine & Kimoto, Yukinori & Knuchel, Dominique & Kogura,
Norikazu & Kurabe, Keita & Mansfield, John & Narrog, Heiko & Pratiwi, De-
sak Putu Eka & van Putten, Saskia & Senge, Chikako & Tykhostup, Olena. This
volume. Language vs. individuals in cross-linguistic corpus typology. In Haig,
Geoffrey & Schnell, Stefan & Seifart, Frank (eds.), Doing corpus-based typology
with spoken language corpora: State of the art (Language Documentation & Conser-
vation special publication 25), 179–232. Honolulu, HI: University of Hawai’i Press.
(https://hdl.handle.net/10125/74661).
Barth, Danielle & Schnell, Stefan. 2021. Understanding corpus linguistics. London: Rout-
ledge.
Behaghel, Otto. 1909. Beziehungen zwischen Umfang und Reihenfolge von
Satzgliedern. Indogermanische Forschungen 25. 110–142. (https://doi.org/10.
1515/9783110242652.110).
Bentz, Christian & Ferrer-i-Cancho, Ramon. 2016. Zipf’s law of abbreviation as a lan-
guage universal. In Bentz, Christian & Jäger, Gerhard & Yanovich, Igot (eds.), Pro-
ceedings of the Leiden Workshop on Capturing Phylogenetic Algorithms for Linguist-
ics. Tübingen: University of Tubingen. (https://publikationen.uni-tuebingen.
de/xmlui/handle/10900/68558).
Berez-Kroeker, Andrea L. & Andreassen, Helene N. & Gawne, Lauren & Holton, Gary
& Kung, Susan Smythe & Pulsifer, Peter & Collister, Lauren B. & The Data
Citation and Attribution in Linguistics Group & The Linguistics Data Interest
Group. 2018. The Austin Principles of Data Citation in Linguistics (Version 1.0): In-
troduction and guidelines for annotators (Version 7.0). (http : / / site . uit . no /
linguisticsdatacitation/austinprinciples/).
20 LD&C SP25 — Doing corpus-based typology with spoken language data

Berman, Ruth A. & Slobin, Dan I. 1994. Relating events in narrative: A cross-linguistic
developmental study. Mahwah, NJ: Erlbaum.
Biber, Douglas. 1995. Dimensions of register variation: A cross-linguistic comparison.
Cambridge: Cambridge University Press.
Bickel, Balthasar. 2003. Referential density in discourse and syntactic typology. Lan-
guage 79(4). 708–736.
Bickel, Balthasar. 2009. Typological patterns and hidden diversity. Plenary talk de-
livered at the 8th Biannual Meeting of the Association for Linguistic Typo-
logy, Berkeley, United States of America, 24 July 2013. (https : / / www .
comparativelinguistics . uzh . ch / dam / jcr : 00000000 - 774a - e877 - 0000 -
0000407727a0/alt2009bickel-plenary.pdf).
Bickel, Balthasar. 2015. Distributional typology. In Heine, Bernd & Narrog, Heiko
(eds.), The Oxford handbook of linguistic analysis, 2nd edn. Oxford: Oxford Uni-
versity Press. (https://doi.org/10.1093/oxfordhb/9780199677078.013.0046).
Blasi, Damian & Cotterell, Ryan & Wolf-Sonkin, Lawrence & Stoll, Sabine & Bickel,
Balthasar & Baroni, Marco. 2019. On the distribution of deep clausal embeddings:
A large cross-linguistic study. In Korhonen, Anna & Traum, David & Màrquez,
Lluís (eds.), Proceedings of the 57th Annual Meeting of the Association for Computa-
tional Linguistics, 3938–3943. Florence: Association for Computational Linguist-
ics.
Bromham, Lindell & Dinnage, Russell & Skirgård, Hedvig & Ritchie, Andrew & Car-
dillo, Marcel & Meakins, Felicity & Greenhill, Simon & Hua, Xia. 2021. Global pre-
dictors of language endangerment and the future of linguistic diversity. Nature
Ecology & Evolution. (https://doi.org/10.1038/s41559-021-01604-y).
Bybee, Joan L. & Pagliuca, William & Perkins, Revere D. 1990. On the asymmetries
in the affixation of grammatical material. In Croft, William & Denning, Keith &
Kemmer, Suzanne (eds.), Studies in typology and diachrony: Papers presented to
Joseph H. Greenberg on his 75th birthday, 1–42. Amsterdam: John Benjamins.
Chafe, Wallace (ed.). 1980. The Pear Stories: Cognitive, cultural, and linguistic aspects
of narrative production. Norwood, NJ: Ablex.
Chanard, Christian. 2015. ELAN-CorpA: Lexicon-aided annotation in ELAN. In
Mettouchi, Amina & Chanard, Christian (eds.), CorpAfroAs: The CorpAfroAs cor-
pus of spoken AfroAsiatic languages. (https : / / doi . org / 10 . 1075 / scl . 68 .
website).
Cohen Priva, Uriel. 2017. Informativity and the actuation of lenition. Language 93(3).
569–597.
Comrie, Bernard. 1981. Language universals and linguistic typology. London: Black-
well.
Language documentation and corpus-based typology 21

Coupé, Christophe & Oh, Yoon & Dediu, Dan & Pellegrino, François. 2019. Different
languages, similar encoding efficiency: Comparable information rates across the
human communicative niche. Science Advances 5(9). eaaw2594. (https : / / doi .
org/10.1126/sciadv.aaw2594).
Cutler, Ann & Hawkins, John A. & Gilligan, Gary. 1985. The suffixing preference: A
processing explanation. Linguistics 23(5). 723–758. (https://doi.org/10.1515/
ling.1985.23.5.723).
Cysouw, Michael. 2014. Inducing semantic roles. In Luraghi, Silvia & Narrog, Heiko
(eds.), Perspectives on Semantic Roles, 23–68. Berlin: Mouton de Gruyter. (https:
//doi.org/10.1075/tsl.106.02cys).
Dahl, Östen. 2015. How WEIRD are WALS languages? Paper presented at the Closing
Conference of the Department of Linguistics at the Max Planck Institute for Evol-
utionary Anthropology, Leipzig, Germany, 1–3 May 2015. (https : / / www . eva .
mpg.de/fileadmin/content_files/linguistics/conferences/2015-diversity-
linguistics/Dahl_slides.pdf).
Dingemanse, Mark & Blythe, Joe & Dirksmeyer, Tyko. 2014. Formats for other-
initiation of repair across languages: An exercise in pragmatic typology. Studies
in Language 38(1). 5–43. (https://doi.org/10.1075/sl.38.1.01din).
Dingemanse, Mark & Rossi, Giovanni & Floyd, Simeon. 2017. Place reference in
story beginnings: A cross-linguistic study of narrative and interactional afford-
ances. Language in Society 46(2). 129–158. (https : / / doi . org / 10 . 1017 /
S0047404516001019).
Donohue, Mark. 1999. A grammar of Tukang Besi. Berlin: Mouton de Gruyter.
Dryer, Matthew S. 1992. The Greenbergian word order correlations. Language 68(1).
81–138.
Dryer, Matthew S. 2013. Feature 101A: Expression of pronominal subjects. In Dryer,
Matthew S. & Haspelmath, Martin (eds.), The World Atlas of Language Structures
Online. Leipzig: Max Planck Institute for Evolutionary Anthropology. (https://
wals.info/feature/101A).
Duranti, Alessandro. 1981. The Samoan fono: A sociolinguistic study. Canberra: Aus-
tralian National University.
Duranti, Alessandro. 1997. Linguistic anthropology. Cambridge: Cambridge University
Press.
ELAN developers. 2020. ELAN (Version 6.2). Nijmegen: Max Planck Institute for Psy-
cholinguistics. (https://archive.mpi.nl/tla/elan).
Frank, Austin F. & Jaeger, Florian. 2008. Speaking rationally: Uniform information
density as an optimal strategy for language production. Proceedings of the 30th An-
nual Meeting of the Cognitive Science Society (CogSci-08), Washington D.C., United
States of America, 23–26 July 2008. 939–944.
22 LD&C SP25 — Doing corpus-based typology with spoken language data

Futrell, Richard & Levy, Roger P. & Gibson, Edward. 2020. Dependency locality as an
explanation principle for word order. Language 76(2). 371–412.
Futrell, Richard & Mahowald, Kyle & Gibson, Edward. 2015. Quantifying word order
freedom in dependency corpora. Proceedings of the 3rd International Conference
on Dependency Linguistics (Depling 2015), Uppsala, Sweden, 24–26 August 2015. 91–
100.
Gerdes, Kim & Kahane, Sylvain & Chen, Xinying. 2021. Typometrics: From implica-
tional to quantitative universals in word order typology. Glossa 6(1). 1–31. (https:
//doi.org/{10.5334/gjgl.764}).
Gibson, Edward & Futrell, Richard & Piantadosi, Steven T. & Dautriche, Isabelle &
Mahowald, Kyle & Bergen, Leon & Levy, Roger. 2019. How efficiency shapes hu-
man language. Trends in Cognitive Sciences 23(12). 1087. (https://doi.org/10.
1016/j.tics.2019.09.005).
Gippert, Jost & Himmelmann, Nikolaus P. & Mosel, Ulrike (eds.). 2006. Essentials of
language documentation. Berlin: Mouton de Gruyter.
Greenberg, Joseph H. 1963. Some universals of grammar with particular reference
to the order of meaningful elements. In Greenberg, Joseph H. (ed.), Universals of
language, 73–113. Cambridge, MA: MIT Press.
Greenberg, Joseph H. 1966. Language universals, with special reference to feature hier-
archies. The Hague: Mouton.
Haig, Geoffrey & Schnell, Stefan. 2014. Annotations using GRAID (Grammatical Rela-
tions and Animacy in Discourse): Introduction and guidelines for annotators (Version
7.0). (https://multicast.aspra.uni-bamberg.de/#annotations).
Haig, Geoffrey & Schnell, Stefan (eds.). 2021. Multi-CAST: Multilingual Corpus of An-
notated Spoken Texts. Version 2108. (https : / / multicast . aspra . uni - bamberg .
de/).
Haig, Geoffrey & Schnell, Stefan & Schiborr, Nils N. This volume. Universals of ref-
erence in discourse and grammar: Evidence from the Multi-CAST collection of
spoken corpora. In Haig, Geoffrey & Schnell, Stefan & Seifart, Frank (eds.), Doing
corpus-based typology with spoken language corpora: State of the art (Language Doc-
umentation & Conservation special publication 25), 141–177. Honolulu, HI: Univer-
sity of Hawai’i Press. (https://hdl.handle.net/10125/74660).
Haig, Geoffrey & Schnell, Stefan & Wegener, Claudia. 2011. Comparing corpora from
endangered languages: Explorations in language typology based on original texts.
In Haig, Geoffrey & Nau, Nicole & Schnell, Stefan & Wegener, Claudia (eds.),
Documenting endangered languages: Achievements and perspectives, 55–86. Berlin:
Mouton de Gruyter. (https://doi.org/10.1515/9783110260021.55).
Language documentation and corpus-based typology 23

Hale, Ken & Krauss, Michael & Watahomigie, Lucille J. & Yamamoto, Akira Y. & Craig,
Colette & Masayesva Jeanne, LaVerne & England, Nora C. 1992. Endangered lan-
guages. Language 68(1). 1–42. (https://doi.org/10.1353/lan.1992.0052).
Haspelmath, Martin. 2021. Explaining grammatical coding asymmetries: Form-
frequency correspondences and predictability. Journal of Linguistics 57(3). 605–
633. (https://doi.org/10.1017/S0022226720000535).
Haspelmath, Martin & Calude, Andreea & Spagnol, Michael & Narrog, Heiko & Bamy-
aci, Elif. 2014. Coding causal-noncausal verb alternations: A form-frequency cor-
respondence explanation. Journal of Linguistics 50(3). 587–625.
Hawkins, John A. 1994. A performance theory of order and constituency. Cambridge:
Cambridge University Press.
Hawkins, John A. 2004. Efficiency and complexity in grammars. Oxford: Oxford Uni-
versity Press.
Hawkins, John A. 2014. Cross-linguistic variation and efficiency. Oxford: Oxford Uni-
versity Press.
Hellwig, Birgit & Defina, Rebecca & Kidd, Evan & Allen, Shanley & Davidson, Lucy &
Kelly, Barbara F. This volume. Child language documentation: The sketch acquis-
ition project. In Haig, Geoffrey & Schnell, Stefan & Seifart, Frank (eds.), Doing
corpus-based typology with spoken language corpora: State of the art (Language
Documentation & Conservation special publication 25), 29–58. Honolulu, HI: Uni-
versity of Hawai’i Press. (https://hdl.handle.net/10125/74657).
Henrich, Joseph & Heine, Steven J. & Norenzayan, Ara. 2010. The weirdest people in
the world? Behavioral and Brain Sciences 33(2–3). 61–83. (https://doi.org/10.
1017/S0140525X0999152X).
Hildebrandt, Kristine A. & Jany, Carmen & Silva, Wilson (eds.). 2017. Documenting
variation in endangered languages (Language Documentation & Conservation spe-
cial publication 13). Honolulu, HI: University of Hawai’i Press. (http : / / hdl .
handle.net/10125/24754).
Himmelmann, Nikolaus P. 1998. Documentary and descriptive linguistics. Linguistics
36(2). 161–195.
Himmelmann, Nikolaus P. 2014. Asymmetries in the prosodic phrasing of function
words: Another look at the suffixing preference. Language 90(4). 927–960.
Hymes, Dell H. 1961. Functions of speech: An evolutionary approach. In Gruber, Fre-
derick C. (ed.), Anthropology and education, 55–83. Philadelphia, PA: University
of Philadelphia Press.
Hymes, Dell H. 1962. The ethnography of speaking. In Gladwin, Thomas & Sturtevant,
William C. (eds.), Anthropology and human behaviour, 13–53. Washington, D.C.:
Anthropological Society of Washington.
24 LD&C SP25 — Doing corpus-based typology with spoken language data

Inbar, Maya & Grossman, Eitan & Landau, Ayelet N. 2020. Sequences of intonation
units form a 1 Hz rhythm. Scientific Reports 10(1). 15846. (https://doi.org/10.
1038/s41598-020-72739-4).
Izre’el, Shlomo & Mettouchi, Amina. 2015. Representation of speech in CorpAfroAs:
Transcriptional strategies and prosodic units. In Mettouchi, Amina & Vanhove,
Martine & Caubet, Dominique (eds.), Corpus-based studies of lesser-described lan-
guages: The CorpAfroAs corpus of spoken AfroAsiatic, 13–41. Amsterdam: John Ben-
jamins.
Jaeggli, Osvaldo & Safir, Kenneth J. 1989. The null subject parameter and parametric
theory. In Jaeggli, Osvaldo & Safir, Kenneth J. (eds.), The null subject parameter,
1–44. Dordrecht: Kluwer.
Jing, Yingqi & Widmer, Paul & Bickel, Balthasar. 2021. Word order variation is partially
constrained by syntactic complexity. Cognitive Science 45(11). e13056. (https://
doi.org/10.1111/cogs.13056).
Just, Erika & Čéplö, Slavomír. To appear. Differential object indexing in Maltese — a
corpus based pilot study. In Turek, Przemysław & Nintemann, Julia (eds.), Maltese:
Contemporary changes and historical innovations. Mouton de Gruyter: Berlin.
Koplenig, Alexander & Meyer, Peter & Wolfer, Sascha & Müller-Spitzer, Carolin. 2017.
The statistical trade-off between word order and word structure — Large-scale
evidence for the principle of least effort. PLoS ONE 12(3). e0173614. (https://doi.
org/10.1371/journal.pone.0173614).
Labov, William. 1972. Sociolinguistic patterns. Philadelphia, PA: University of
Pennsylvania Press.
Labov, William. 1994. Principles of linguistic change: Internal factors. Malden, MA:
Blackwell.
Levinson, Stephen C. & Evans, Nicholas. 2010. Time for a sea change in linguistics:
Response to comments on ‘The myth of language universals’. Lingua 12. 2733–
2758. (https://doi.org/10.1016/j.lingua.2010.08.001).
Levshina, Natalia. 2019. Token-based typology and word order entropy: A study based
on universal dependencies. Languages in Contrast 23(3). 533–572. (https://doi.
org/10.1515/lingty-2019-0025).
Levshina, Natalia. 2021a. Corpus-based typology: Applications, challenges and some
solutions. Linguistic Typology. (https://doi.org/10.1515/lingty-2020-0118).
Levshina, Natalia. 2021b. Cross-linguistic trade-offs and causal relationships between
cues to grammatical subject and object, and the problem of efficiency-related ex-
planations: A reverse-engineering approach. Frontiers in Psychology 12. 648200.
(https://doi.org/10.3389/fpsyg.2021.648200).
MacWhinney, Brian. 2000. The CHILDES Project: Tools for analysing talk. Mahwah, NJ:
Erlbaum.
Language documentation and corpus-based typology 25

Mansfield, John & Stanford, James N. 2017. Documenting sociolinguistic variation

in lesser-studied indigenous communities: Challenges and practical solutions. In
Hildebrandt, Kristine A. & Jany, Carmen & Silva, Wilson (eds.), Documenting vari-
ation in endangered languages (Language Documentation & Conservation special
publication 13), 116–136. Honolulu, HI: University of Hawai’i Press. (http://hdl.
handle.net/10125/24751).
Matić, Dejan & Wedgwood, Daniel. 2013. The meanings of focus: The significance of
an interpretation-based category in cross-linguistic analysis. Journal of Linguistics
49(1). 127–163. (https://doi.org/10.1017/S0022226712000345).
Mayer, Mercer. 1969. Frog, where are you? New York: Dial Books for Young Readers.
Mayer, Thomas & Cysouw, Michael. 2014. Creating a massively parallel Bible corpus.
Proceedings of the 9th International Conference on Language Resources and Evalu-
ation (LREC’14), Reykjavik, Iceland, 26–31 May 2014.
McEnery, Tony & Wilson, Andrew. 2001. Corpus linguistics: An introduction. Edin-
burgh: Edinburgh University Press.
Mettouchi, Amina & Vanhove, Martine. This volume. Prosodic segmentation and
cross-linguistic comparison in CorpAfroAs and CorTypo: Corpus-driven and
corpus-based approaches. In Haig, Geoffrey & Schnell, Stefan & Seifart, Frank
(eds.), Doing corpus-based typology with spoken language corpora: State of the art
(Language Documentation & Conservation special publication 25), 59–113. Hon-
olulu, HI: University of Hawai’i Press. (https://hdl.handle.net/10125/74658).
Meyerhoff, Miriam. 2009. Replication, transfer, and calquing: Using variation as a tool
in the study of language contact. Language Variation and Change 21(3). 297–317.
Moran, Steven & Schikowski, Robert & Pajović, Danica & Hysi, Cazim & Stoll, Sabine.
2016. The ACQDIV database: Min(d)ing the ambient language. Proceedings of the
10th Conference on Language Resources and Evaluation (LREC’16), Portorož, Slove-
nia, 23–28 May 2020. 4423–4429.
Newman, Paul. 2013. The law of unintended consequences: How the endangered lan-
guages movement undermines field linguistics as a scientific enterprise. Seminar
talk delivered at the School of Orential and African Studies, University of Lon-
don, London, United Kingdom, 15 October 2013.
Ozerov, Pavel. This volume. This research topic of yours – is it a research topic at all?
Using comparative interactional data for a fine-grained reanalysis of traditional
concepts. In Haig, Geoffrey & Schnell, Stefan & Seifart, Frank (eds.), Doing corpus-
based typology with spoken language corpora: State of the art (Language Document-
ation & Conservation special publication 25), 233–280. Honolulu, HI: University of
Hawai’i Press. (https://hdl.handle.net/10125/74662).
26 LD&C SP25 — Doing corpus-based typology with spoken language data

Piantadosi, Steven T. & Tily, Harry J. & Gibson, Edward. 2011. Word lengths are optim-
ized for efficient communication. Proceedings of the National Academy of Sciences
108(9). 3526–3529. (https://doi.org/10.1073/pnas.1012551108).
Piantadosi, Steven T. & Tily, Harry J. & Gibson, Edward. 2012. The communicative
function of ambiguity in language. Cognition 122(3). 1280–1291.
Pimentel, Tiago & Meister, Clara & Salesky, Elizabeth & Teufel, Simone & Blasi, Dam-
ián & Cotterell, Ryan. 2021. A surprisal–duration trade-off across and within the
world’s languages. Preprint published on arXiv 2109.15000. (http://arxiv.org/
abs/2109.15000).
Rizzi, Luigi. 1982. Issues in Italian syntax. Dordrecht: Foris.
Roberts, Ian & Holmberg, Anders. 2009. Introduction: Parameters in minimalist theory.
In Biberauer, Theresa & Holmberg, Anders & Roberts, Ian & Sheehan, Michelle
(eds.), Parametric variation: Null subjects in minimalist theory, 1–57. Cambridge:
Cambridge University Press.
Sacks, Harvey & Schegloff, Emanuel A. & Jefferson, Gail. 1974. A simplest systemat-
ics for the organization of turn-taking in conversation. Language 50(4). 696–735.
(https://doi.org/10.2307/412243).
Salesky, Elizabeth & Chodroff, Eleanor & Pimentel, Tiago & Wiesner, Matthew & Cot-
terell, Ryan & Black, Alan W. & Eisner, Jason. 2020. A corpus for large-scale phon-
etic typology. Proceedings of the 58nd Annual Meeting of the Association for Com-
putational Linguistics (ACL’20). 4526–4546. (https : / / doi . org / 10 . 18653 / v1 /
2020.acl-main.415).
Schegloff, Emanuel A. 2006. Sequence organization in interaction: A primer in conver-
sation analysis. Cambridge: Cambridge University Press.
Schnell, Stefan & Barth, Danielle. 2018. Discourse motivations for pronominal and
zero objects across registers in Vera’a. Language Variation and Change 30(1). 51–
81.
Schnell, Stefan & Schiborr, Nils N. In press. Cross-linguistic corpus studies in linguistic
typology. Annual Review of Linguistics.
Seifart, Frank. This volume. Combining documentary linguistics and corpus phonetics
to advance corpus-based typology. In Haig, Geoffrey & Schnell, Stefan & Seifart,
Frank (eds.), Doing corpus-based typology with spoken language corpora: State of
the art (Language Documentation & Conservation special publication 25), 115–139.
Honolulu, HI: University of Hawai’i Press. (https : / / hdl . handle . net / 10125 /
74659).
Seifart, Frank & Evans, Nicholas & Hammarström, Harald & Levinson, Stephen C.
2018. Language documentation 25 years on. Language 94(4). e324–e345. (https:
//doi.org/10.1353/lan.2018.0070).
Language documentation and corpus-based typology 27

Seifart, Frank & Haig, Geoffrey & Himmelmann, Nikolaus P. & Jung, Dagmar & Mar-
getts, Anne & Trilsbeek, Paul (eds.). 2012. Potentials of language documentation:
Methods, analyses, and utilization (Language Documentation & Conservation spe-
cial publication 3). Honolulu, HI: University of Hawai’i Press. (http : / / hdl .
handle.net/10125/4510).
Slobin, Dan I. (ed.). 1985. The crosslinguistic study of language acquisition, Volume 1:
The data. Mahwah, NJ: Erlbaum.
Stanford, James N. & Preston, Dennis R. (eds.). 2009. Variation in indigenous and minor-
ity languages. Amsterdam: John Benjamins.
Stoll, Sabine & Bickel, Balthasar. 2009. How deep are differences in referential dens-
ity? In Guo, Jiansheng & Lieven, Elena & Budwig, Nancy & Ervin-Tripp, Susan &
Nakamura, Keiko & Özçalışkan, Şeyda (eds.), Crosslinguistic approaches to the psy-
chology of language: Research in the tradition of Dan Isaac Slobin, 543–555. London:
Psychology Press.
Stoll, Sabine & Bickel, Balthasar. 2013. Capturing diversity in language acquisition
research. In Bickel, Balthasar & Grenoble, Lenore A. & Peterson, David A. & Tim-
berlake, Alan (eds.), Language typology and historical contingency, 195–216. Ams-
terdam: Benjamins.
Thieberger, Nicholas & Berez, Andrea L. 1963. Linguistic data management. In
Thieberger, Nicholas (ed.), The Oxford handbook of linguistic fieldwork, 90–118.
Oxford: Oxford University Press.
Tomlin, Russell. 1986. Basic word order: Functional principles. London: Routledge.
Torres Cacoullos, Rena & Travis, Catherine E. 2018. Bilingualism in the community:
Code-switching and grammars in contact. Cambridge: Cambridge University Press.
(https://doi.org/10.1017/9781108235259).
Torres Cacoullos, Rena & Travis, Catherine E. 2019. Variationist typology: Shared
probabilistic constraints across (non-)null subject languages. Linguistics 57(3).
653–692.
Vollmer, Maria C. 2019. How radical is pro-drop in Mandarin? A quantitative corpus
study on referential choice in Mandarin Chinese. MA thesis, University of Bamberg.
Wälchli, Bernhard. 2009. Data reduction typology and the bimodal distribution bias.
Linguistic Typology 13(1). 77–94.
Wälchli, Bernhard & Sölling, Arnd. 2013. The encoding of motion events: Building
typology bottom-up from text data in many languages. In Goschler, Juliana &
Stefanowitsch, Anatol (eds.), Variation and change in the encoding of motion events,
102–125. Amsterdam: John Benjamins.
Wilkinson, Mark D. & Dumontier, Michel & Aalbersberg, IJsbrand J. & Appleton, Gab-
rielle & Axton, Myles & Baak, Arie & Blomberg, Niklas & alii. 2016. The FAIR Guid-
28 LD&C SP25 — Doing corpus-based typology with spoken language data

ing Principles for scientific data management and stewardship. Scientific Data 3(1).
1–9. (https://doi.org/10.1038/sdata.2016.18).
Zakharko, Taras & Witzlack-Makarevich, Alena & Nichols, Johanna & Bickel,
Balthasar. 2017. Late aggregation as a design principle for typological databases.
Paper presented at the ALT Workshop on Design Principles of Typological Data-
bases, Canberra, Australia, 15 December 2017.
Zeman, Daniel & Nivre, Joakim & Abrams, Mitchell & alii. 2021. Universal Dependen-
cies 2.8. Prague: Universal Dependencies Consortium. (https://hdl.handle.net/
11234/1-3687).
Zipf, George K. 1935. The psycho-biology of Language: An introduction to dynamic philo-
logy. Cambridge, MA: MIT.